A system and method for identifying an anesthetic drug

Abstract

The application discloses a narcotic drug identification system and method, and relates to the technical field of identification, and solves the technical problems that when the complex and various narcotic drug identification is carried out, the identification is still not accurate, new drugs cannot be processed, and the abnormal identification of non-new drugs is not handled well. Through pre-processing of the collected image and feature extraction by combining a deep learning algorithm, the shape, color, texture and character of the narcotic drug identification can be more accurately obtained, the accuracy of comparison with standard identification features in a database is improved, the misjudgment rate is reduced, for the abnormal identification of non-new drugs, the correlation between drug features is analyzed in depth, the correlation features are screened out, the redundant features are removed, the same type of product features are automatically learned and fused by using a deep learning model, more representative fusion features are formed for re-identification, and the processing capacity for the abnormal identification of non-new drugs is improved.

Description

Technical Field

[0001] This invention relates to the field of identification technology, specifically to an identification system and method for anesthetic drugs. Background Technology

[0002] With the development of the medical industry, anesthetic drugs are widely used in clinical treatment. Accurate identification of anesthetic drug labels is crucial for ensuring medication safety and standardizing drug management.

[0003] According to patent publication number CN116932539A, this patent discloses a system for identifying narcotic drugs, which includes the following steps: Step 1, building a tree model; Step 2, entering the rule tree from the root entry point, first checking whether the identifier is already contained in the database using a hash mapping algorithm. If it exists, the result is returned directly; otherwise, proceed to Step 3; Step 3, rule sorting; Step 4, rule matching; Step 5, entering the sub-rule tree of rule1, repeating Steps 3 and 4, iterating in a loop. If the match is successful, the matching result is returned; if it fails, the identifier is returned to the previous level. This invention can analyze drug identification standards, extract the features of the standards, abstract the features into paradigms, then construct a tree-shaped matching model to reduce matching redundancy, then prioritize each level using information content, and finally use a tree-shaped matching algorithm for matching and identification. It has the advantages of high recognition rate and fast speed.

[0004] Traditional drug labeling and identification methods rely heavily on manual visual judgment, which is inefficient and prone to misjudgment. With the rise of image recognition technology, some image recognition-based drug labeling and identification systems have begun to be applied. However, when faced with complex and diverse anesthetic drug labels, problems still exist, such as inaccurate identification, inability to handle new drugs, and insufficient handling of abnormal identification situations for non-new drugs. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a system and method for identifying anesthetic drugs, which solves the problems of inaccurate identification, inability to handle novel drugs, and insufficient response to abnormal identification situations of non-novel drugs when identifying complex and diverse anesthetic drugs.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a system for identifying and labeling narcotic drugs, comprising:

[0007] The database comparison and analysis module is used to preprocess the images of the anesthetic drugs themselves transmitted by the target image acquisition module to obtain preprocessed images, and to extract the corresponding image features using deep learning algorithms. At the same time, it compares the images with the database to generate a matching result indicating whether the images exist or not.

[0008] The matching success analysis module is used to analyze the matching results and generate single or compound matching signals based on the number of matching results. For a single matching signal, identification information or secondary analysis signals are generated by comparing image features with drug labels.

[0009] The secondary analysis signal is processed to generate new or non-new drug signals for the current drug classification, and at the same time, identification information or identification anomaly information is generated by combining historical identification information in historical data.

[0010] For composite matching signals, the recognition features of the matching results are obtained and matched with image features to generate recognition information;

[0011] The comparison failure analysis module is used to analyze the absence of matching results, generate new drug signals or non-new drug signals for the current drug classification, obtain the corresponding drug information to generate identification information for new drug signals, analyze non-new drug signals, obtain drug features and filter related features, and combine them with drug features of similar products to obtain fused features, generate identification information based on fused features, and transmit it to the identification information output module.

[0012] As a further embodiment of the present invention, it also includes a target image acquisition module and a recognition information output module;

[0013] The target image acquisition module is used to acquire images of the anesthetic drugs themselves through image acquisition equipment and transmit them to the database comparison and analysis module;

[0014] The identification information output module is used to display the acquired identification information and identification anomaly information to the corresponding operators.

[0015] As a further aspect of the present invention, the database comparison and analysis module generates the existence or non-existence of matching results in the following specific manner:

[0016] The acquired image is preprocessed by performing grayscale conversion, noise reduction, filtering, and binarization. Then, deep learning algorithms in image recognition software are used to extract features from the preprocessed image to obtain image features.

[0017] The obtained image features are compared and matched with the standard identifier features in the database to generate matching results. If the image features exist in the database, a matching result is generated and transmitted to the matching success analysis module; otherwise, a matching result is generated and transmitted to the matching failure analysis module.

[0018] As a further aspect of the present invention, the specific method by which the comparison success analysis module analyzes the matching results is as follows:

[0019] Obtain the matching results and determine the number of matching results. If there is only one set of matching results, generate a single matching signal. If there are multiple sets of matching results, generate a composite matching signal and analyze them separately.

[0020] As a further aspect of the present invention, the specific method by which the matching success analysis module analyzes a single matching signal is as follows:

[0021] Obtain a single matching result and its corresponding drug identifier, compare the image features with the drug identifier, if they match, the identifier is determined to be correctly identified, generate identification information and transmit it to the identification information output module; if they do not match, the identifier is identified abnormally, generate a secondary analysis signal and carry out subsequent analysis.

[0022] Analyze drug production information to determine if it is a new drug, and generate a new drug signal or a non-new drug signal. If it is a new drug, enter the drug identification into the database and process it to generate identification information. If it is a non-new drug, search historical identification data to see if there are any cases that are the same as the current drug and cannot be identified. If they exist, obtain the corresponding processing results and generate identification information. If they do not exist, generate identification anomaly information.

[0023] As a further aspect of the present invention, the specific method by which the matching success analysis module analyzes the composite matching signal is as follows:

[0024] All matching results are obtained, and the corresponding recognition features are acquired. The recognition features are then matched with the image features of the current drug. Successful matching results are recorded as results to be analyzed. Then, based on the information corresponding to the results to be analyzed, recognition information is generated and transmitted to the recognition information output module.

[0025] As a further aspect of the present invention, the specific method by which the comparison failure analysis module analyzes the absence of matching results is as follows:

[0026] Determine whether the current drug is a novel drug, generate a novel drug signal or a non-novel drug signal, and analyze the two separately;

[0027] The process involves analyzing the generated new drug signals, obtaining the drug information corresponding to the current drug, storing the drug information in the database for updating, generating identification information, and then transmitting it to the identification information output module.

[0028] As a further aspect of the present invention, the specific method by which the comparison failure analysis module analyzes the signals of non-novel drugs is as follows:

[0029] Based on the acquired basic drug information, the drug characteristics corresponding to non-new drugs are acquired and labeled as i, where i = 1, 2, ..., j, and j represents the type of drug characteristic. Then, the correlation between drug characteristics is analyzed, and related characteristics are selected. Related characteristics are retained, and the remaining drug characteristics are removed.

[0030] Next, the drug characteristics of similar products are obtained and recorded as similar characteristics. At the same time, the similar characteristics are fused with the obtained associated characteristics to obtain fused characteristics. Based on the fused characteristics and the database, identification information is generated and then transmitted to the identification information output module.

[0031] A method for identifying narcotic drugs, comprising:

[0032] The image of the anesthetic drug itself is acquired and preprocessed to obtain a preprocessed image. Image features are extracted based on a convolutional neural network. The image features are compared with the database to generate matching results and non-matching results.

[0033] For existing matching results analysis, a single matching signal or a composite matching signal is generated based on the number of matching results. The matching results corresponding to the single matching signal are identified and matched to generate identification information or secondary analysis signals.

[0034] For secondary analysis signal processing, new drug signals or non-new drug signals are generated based on production conditions, and at the same time, they are matched and compared with the database to generate identification information and identification anomaly information.

[0035] For composite matching signal analysis, recognition information is generated by matching the recognition features of the matching results with image features.

[0036] For cases where no matching results are found, a new drug signal or a non-new drug signal is generated for the current drug classification. For new drug signals, the corresponding drug information is obtained to generate identification information.

[0037] For non-novel drug signal analysis, drug characteristics are acquired and related features are screened. At the same time, drug characteristics of similar products are fused to obtain fused features. Identification information is generated based on the fused features.

[0038] This invention provides a system and method for identifying narcotic drugs. Compared with the prior art, it has the following advantages:

[0039] This invention preprocesses acquired images through grayscale conversion, noise reduction, filtering, and binarization, and then combines this with deep learning algorithms for feature extraction. This enables more accurate acquisition of features such as shape, color, texture, and characters of narcotic drug identifiers, improving the accuracy of comparison with standard identifier features in the database and reducing the false positive rate. The system has the function of automatically determining whether a drug is a novel drug. If it is a novel drug, it can automatically collect drug information and enter it into the database, while generating comprehensive identification information to provide data support for subsequent identification, enabling the system to adapt promptly to the emergence of new drugs on the market.

[0040] For anomalies in the identification of non-novel drugs, the system can deeply analyze the correlation between drug features based on basic information such as drug appearance, text, color, and shape, filter out related features, and eliminate redundant features. It also uses a deep learning model to automatically learn and fuse features of similar products to form more representative fused features for re-identification. Furthermore, by leveraging historical identification data, if similar anomalies exist, the system can directly obtain the corresponding processing results to generate identification information, thus improving its ability to handle anomalies in the identification of non-novel drugs. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the system of the present invention;

[0042] Figure 2 This is a diagram illustrating the steps and methods of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example 1, please refer to Figure 1 This application provides a system for identifying narcotic drugs, including: a target image acquisition module, a database comparison and analysis module, a comparison success analysis module, a comparison failure analysis module, and an identification information output module, and in combination with... Figure 1 It can be seen that the above functional modules are connected electrically in one direction.

[0045] The target image acquisition module acquires images of the anesthetic drug itself using an image acquisition device, such as a high-definition camera. The image itself specifically includes a picture of the entire anesthetic drug. Simultaneously, the acquired image is transmitted to a database comparison and analysis module.

[0046] The database comparison and analysis module is used to preprocess the acquired images to obtain preprocessed images. Preprocessing operations include grayscale conversion (converting a color image to a grayscale image to remove color interference and simplify image information; for example, converting a common color medicine box image to grayscale to highlight the outlines of patterns and text), noise reduction (reducing noise points in the image using specific algorithms), filtering (further smoothing the image and improving its clarity and details; for example, Gaussian filtering adjusts pixel values ​​based on the weighted average of the surrounding neighborhood, making image edges smoother), and binarization (converting the image to only black and white to highlight key information; for text and patterns on medicine labels, binarization makes their shape and outline clearer). Then, deep learning algorithms in the image recognition software are used... For example, a convolutional neural network (CNN) extracts features from a preprocessed image to obtain image features, which include shape, color, texture, and characters. For instance, for an injection bottle of an anesthetic drug, the shape feature of the bottle label might be a specific rectangular label shape, the color feature might be a blue background with white text, the texture feature might be the subtle printing texture on the label surface, and the character feature might be textual information such as the drug name, specifications, and production batch number. The obtained image features are compared and matched with standard identifier features in a database, where the identifier features are those corresponding to the drug itself. Matching results are generated, including those with and without matching results. For those with matching results, the results are transmitted to the matching success analysis module; for those without matching results, the results are transmitted to the matching failure analysis module.

[0047] The matching success analysis module is mainly responsible for conducting in-depth analysis of the acquired matching results and generating corresponding signals based on the number of matching results, namely, a single matching signal or a composite matching signal. Specifically, when the system detects that there is only one set of matching results, the module will generate a single matching signal; if there are multiple sets of matching results, a composite matching signal will be generated. Then, detailed analysis will be performed on these two types of signals respectively.

[0048] Upon receiving a single matching signal, the module first obtains the corresponding single matching result. Taking the identification of "fentanyl transdermal patch" in a hospital pharmacy as an example, if the system finds only one matching result, it will obtain the drug identifier corresponding to that matching result. Then, it will carefully compare the previously extracted image features, such as the shape, color, texture, and characters on the patch packaging, with the obtained drug identifier.

[0049] If the image features perfectly match the drug label—for example, the patch packaging is a specific rectangle, the color is blue, the texture shows regular printing patterns, and the character information (drug name, specifications, manufacturer, etc.) also matches precisely—then the label recognition can be determined to be correct. At this point, the module will generate label recognition information containing detailed drug information, such as the drug name "Fentanyl Transdermal Patch," the specification "2.5mg," and the manufacturer "XX Pharmaceutical Co., Ltd."—and transmit this label recognition information to the recognition information output module for subsequent display or recording.

[0050] Conversely, if there are differences between the image features and the drug label, such as the font of the characters in the image being slightly different from the standard label, or the slight color deviation exceeding the allowable range, the system will determine that the label recognition is abnormal and generate a secondary analysis signal.

[0051] Once a secondary analysis signal is generated, the module will immediately analyze the current production status of the drug to determine whether it is a new drug, and then generate a new drug signal or a non-new drug signal, and process the two signals respectively.

[0052] If a new drug signal is generated, it means that the drug may be a newly launched product on the market, and its identification has not yet been fully included in the database. For example, if a pharmaceutical company develops a new anesthetic and analgesic injection, and a hospital purchases it for the first time, the module will record the current drug identification details into the database, including the drug's packaging image, all text information on the label, color features, shape features, etc. After the data is entered, the identification is processed accordingly, such as creating an index for quick retrieval and generating comprehensive identification information, covering the drug's basic attributes, research and development background, etc., providing data support for subsequent identification work.

[0053] When a non-novel drug signal is generated, the module retrieves historical recognition data to check if there are any cases where the same drug could not be recognized previously. For example, if a hospital encounters an identification anomaly when recognizing "ketamine hydrochloride injection," the system, after querying historical data, discovers that a batch of this drug previously had recognition difficulties due to packaging printing issues. If such a similar case is found, the module obtains the corresponding processing results, namely the specific solutions previously developed for the problem and the corresponding recognition results. This might involve adjusting character recognition parameters in the image recognition algorithm or recalibrating the color recognition range. Based on these processing results, the current drug's identification information is generated.

[0054] If no similar cases are found in the historical identification data, an identification anomaly message is generated, prompting staff to conduct further manual verification, such as checking whether the medicine has damaged packaging, tampered labeling, or other anomalies.

[0055] The composite matching signal is analyzed to obtain all matching results. At the same time, the recognition features corresponding to the matching results are obtained and matched with the image features of the current drug. The successfully matched results are recorded as the results to be analyzed. Then, the recognition information is generated based on the information corresponding to the results to be analyzed and transmitted to the recognition information output module.

[0056] The identification information output module is used to display identification information and identification anomaly information to the corresponding operators.

[0057] In this embodiment, as a second embodiment of the present invention, it is implemented based on the first embodiment, and the difference from the first embodiment is as follows:

[0058] The comparison failure analysis module is used to analyze the acquired non-matching results, determine whether the current drug is a new drug, and generate a new drug signal or a non-new drug signal, and analyze both at the same time.

[0059] The system analyzes newly generated drug signals to obtain corresponding drug information, stores this information in a database for updates, and generates identification information, which is then transmitted to the identification information output module. Once a new drug signal is generated, the module immediately begins collecting comprehensive drug information. For example, if a newly developed anesthetic drug, "XX Anesthetic Aerosol," enters the market, the module will obtain information such as its drug name, ingredients, indications, contraindications, dosage, packaging specifications, manufacturer, and approval number. After collection, this drug information is stored in the database and updated regularly to expand the information reserves of the drug identification system. After information storage and updating are complete, the module generates identification information encompassing all the above information and transmits it to the identification information output module, enabling relevant personnel to obtain detailed information about this new drug and providing a basis for subsequent drug management, use, and re-identification.

[0060] The generated non-novel drug signals are analyzed based on the acquired basic drug information, which specifically includes information such as drug appearance, text, color, and shape. At the same time, the drug features corresponding to the non-novel drugs are acquired and labeled as i, where i = 1, 2, ..., j, and j represents the type of drug feature. Then, the correlation between drug features is analyzed, which can be performed using correlation analysis. Correlation features are then selected and retained, while the remaining drug features are removed.

[0061] When a non-novel drug signal is generated, the module operates based on the acquired basic drug information. This basic information includes visually apparent details such as the drug's appearance, text, color, and shape. For example, for "Lidocaine Hydrochloride Injection," the drug appears as a colorless, transparent liquid packaged in ampoules. The ampoules are labeled with the drug name, specifications (e.g., "5ml:0.1g"), manufacturer, and other textual information. The drug is colorless and has a long, thin ampoule shape. Simultaneously, the module acquires various drug characteristics corresponding to this non-novel drug, sequentially labeling them as i, where i = 1, 2, ..., j, and j represents the type of drug characteristic. These characteristics may include the drug's chemical composition, physical properties, and pharmacological effects. For example, in terms of chemical composition, the main component of Lidocaine Hydrochloride Injection is lidocaine hydrochloride, and excipients may include sodium chloride; physical properties include density and pH value; and pharmacological effects include the mechanism of local anesthetic action.

[0062] Next, the module will use correlation analysis to conduct an in-depth analysis of the relationships between drug characteristics. Taking lidocaine hydrochloride injection as an example, the analysis revealed a strong positive correlation between its concentration and the duration of anesthetic effect; that is, the higher the concentration, the longer the duration of anesthetic effect. After analysis and screening, features with significant correlations, such as concentration and duration of anesthetic effect, were identified. These correlated features were retained, while drug characteristics with weak correlations were removed.

[0063] Next, the drug characteristics of similar products are obtained and recorded as similar characteristics. At the same time, the similar characteristics are fused with the obtained associated characteristics to obtain fused characteristics. Based on the fused characteristics and the database, identification information is generated and then transmitted to the identification information output module.

[0064] Subsequently, the module acquires the drug characteristics of similar products and records them as similar characteristics. For lidocaine hydrochloride injection, similar products may include lidocaine hydrochloride injections of different specifications produced by other manufacturers, or other drugs with similar local anesthetic effects, such as procaine hydrochloride injection. Their characteristics are acquired, such as the differences in concentration, packaging specifications, and price between lidocaine hydrochloride injections produced by different manufacturers, and the differences in composition, mechanism of action, and adverse reactions between procaine hydrochloride injections. Next, the similar characteristics are fused with the previously obtained associated characteristics to obtain fused characteristics. For example, the concentration characteristics of lidocaine hydrochloride injections from different manufacturers and the mechanism of action characteristics of procaine hydrochloride injections are integrated with the previously screened correlation characteristics between the concentration and duration of anesthetic effect of lidocaine hydrochloride injections to form a more comprehensive and representative fusion feature set. The feature fusion method here uses a deep learning model to automatically learn feature fusion techniques. For instance, similar and related features can be used as different input channels and fed into deep learning models such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs). During training, the model automatically learns how to fuse these features to achieve better recognition results. Finally, the fused features are compared with the database for further recognition, generating identification information, which is then transmitted to the identification information output module. The aim is to successfully identify drugs that previously failed to be matched in this way.

[0065] The identification information output module is used to display the acquired identification information to the corresponding operator.

[0066] Example 3, please refer to Figure 2 This application provides a method for identifying narcotic drugs, which specifically includes the following steps:

[0067] The image of the anesthetic drug itself is acquired and preprocessed to obtain a preprocessed image. Image features are extracted based on a convolutional neural network. The image features are compared with the database to generate matching results and non-matching results.

[0068] For existing matching results analysis, a single matching signal or a composite matching signal is generated based on the number of matching results. The matching results corresponding to the single matching signal are identified and matched to generate identification information or secondary analysis signals.

[0069] For secondary analysis signal processing, new drug signals or non-new drug signals are generated based on production conditions, and at the same time, they are matched and compared with the database to generate identification information and identification anomaly information.

[0070] For composite matching signal analysis, recognition information is generated by matching the recognition features of the matching results with image features.

[0071] For cases where no matching results are found, a new drug signal or a non-new drug signal is generated for the current drug classification. For new drug signals, the corresponding drug information is obtained to generate identification information.

[0072] For non-novel drug signal analysis, drug characteristics are acquired and related features are screened. At the same time, drug characteristics of similar products are fused to obtain fused features. Identification information is generated based on the fused features.

[0073] The data in the above formulas are all calculated using numerical values, without substituting the units of the parameters. In addition, the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0074] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A system for identifying and labeling narcotic drugs, characterized in that, include: The database comparison and analysis module is used to preprocess the images of the anesthetic drugs themselves transmitted by the target image acquisition module to obtain preprocessed images, and to extract the corresponding image features using deep learning algorithms. At the same time, it compares the images with the database to generate a matching result indicating whether the images exist or not. The matching success analysis module is used to analyze the matching results and generate single or compound matching signals based on the number of matching results. For a single matching signal, identification information or secondary analysis signals are generated by comparing image features with drug labels. The secondary analysis signal is processed to generate new or non-new drug signals for the current drug classification, and at the same time, identification information or identification anomaly information is generated by combining historical identification information in historical data. For composite matching signals, the recognition features of the matching results are obtained and matched with image features to generate recognition information; The comparison failure analysis module is used to analyze the absence of matching results, generate new drug signals or non-new drug signals for the current drug classification, obtain the corresponding drug information to generate identification information for new drug signals, analyze non-new drug signals, obtain drug features and filter related features, and combine them with drug features of similar products to obtain fused features, generate identification information based on fused features, and transmit it to the identification information output module.

2. The identification system for anesthetic drugs according to claim 1, characterized in that, It also includes a target image acquisition module and a recognition information output module; The target image acquisition module is used to acquire images of the anesthetic drugs themselves through image acquisition equipment and transmit them to the database comparison and analysis module; The identification information output module is used to display the acquired identification information and identification anomaly information to the corresponding operators.

3. The identification system for anesthetic drugs according to claim 1, characterized in that, The database comparison and analysis module generates the existence or non-existence of matching results in the following specific way: The acquired image is preprocessed by performing grayscale conversion, noise reduction, filtering, and binarization. Then, deep learning algorithms in image recognition software are used to extract features from the preprocessed image to obtain image features. The obtained image features are compared and matched with the standard identifier features in the database to generate matching results. If the image features exist in the database, a matching result is generated and transmitted to the matching success analysis module; otherwise, a matching result is generated and transmitted to the matching failure analysis module.

4. The identification system for anesthetic drugs according to claim 1, characterized in that, The specific method by which the matching success analysis module analyzes matching results is as follows: Obtain the matching results and determine the number of matching results. If there is only one set of matching results, generate a single matching signal. If there are multiple sets of matching results, generate a composite matching signal and analyze them separately.

5. The identification system for anesthetic drugs according to claim 4, characterized in that, The specific method by which the matching success analysis module analyzes a single matching signal is as follows: Obtain a single matching result and its corresponding drug identifier, compare the image features with the drug identifier, if they match, the identifier is determined to be correctly identified, generate identification information and transmit it to the identification information output module; if they do not match, the identifier is identified abnormally, generate a secondary analysis signal and carry out subsequent analysis. Analyze drug production information to determine whether it is a new drug, generate a new drug signal or a non-new drug signal. If it is a new drug, enter the drug identification into the database and process it to generate identification information. If it is not a new drug, search historical identification data to see if there are any cases that are the same as the current drug and cannot be identified. If they exist, obtain the corresponding processing results and generate identification information. If they do not exist, generate identification anomaly information.

6. The identification system for anesthetic drugs according to claim 4, characterized in that, The specific method by which the matching success analysis module analyzes the composite matching signal is as follows: All matching results are obtained, and the corresponding recognition features are acquired. The recognition features are then matched with the image features of the current drug. Successful matching results are recorded as results to be analyzed. Then, based on the information corresponding to the results to be analyzed, recognition information is generated and transmitted to the recognition information output module.

7. The identification system for anesthetic drugs according to claim 1, characterized in that, The specific method by which the comparison failure analysis module analyzes cases where no matching results exist is as follows: Determine whether the current drug is a novel drug, generate a novel drug signal or a non-novel drug signal, and analyze the two separately; The process involves analyzing the generated new drug signals, obtaining the drug information corresponding to the current drug, storing the drug information in the database for updating, generating identification information, and then transmitting it to the identification information output module.

8. The identification system for anesthetic drugs according to claim 7, characterized in that, The specific method by which the comparison failure analysis module analyzes signals from non-novel drugs is as follows: Based on the acquired basic drug information, the drug characteristics corresponding to non-new drugs are acquired and labeled as i, where i = 1, 2, ..., j, and j represents the type of drug characteristic. Then, the correlation between drug characteristics is analyzed, and related characteristics are selected. Related characteristics are retained, and the remaining drug characteristics are removed. Next, the drug characteristics of similar products are obtained and recorded as similar characteristics. At the same time, the similar characteristics are fused with the obtained associated characteristics to obtain fused characteristics. Based on the fused characteristics and the database, identification information is generated and then transmitted to the identification information output module.

9. A method for identifying narcotic drugs, executed by an identification system for narcotic drugs according to any one of claims 1-8, characterized in that, include: The image of the anesthetic drug itself is acquired and preprocessed to obtain a preprocessed image. Image features are extracted based on a convolutional neural network. The image features are compared with the database to generate matching results and non-matching results. For existing matching results analysis, a single matching signal or a composite matching signal is generated based on the number of matching results. The matching results corresponding to the single matching signal are identified and matched to generate identification information or secondary analysis signals. For secondary analysis signal processing, new drug signals or non-new drug signals are generated based on production conditions, and at the same time, they are matched and compared with the database to generate identification information and identification anomaly information. For composite matching signal analysis, recognition information is generated by matching the recognition features of the matching results with image features. For cases where no matching results are found, a new drug signal or a non-new drug signal is generated for the current drug classification. For new drug signals, the corresponding drug information is obtained to generate identification information. For non-novel drug signal analysis, drug characteristics are acquired and related features are screened. At the same time, drug characteristics of similar products are fused to obtain fused features. Identification information is generated based on the fused features.

Images